| This dissertation describes the development of a method for visualizing uncertainty and a complimentary framework that can help product development decision makers address three challenges: thinking probabilistically, combining qualitative and quantitative information, and exposing underlying assumptions. Density Clouds, a scatterplot-based method, affords decision makers a visual means of translating their predictions into probability density functions. Decision Analytical Scorecarding provides the framework for incorporating this method into the new product introduction process at an organization.;The need to manage uncertainty grows from recent price volatility, an accepted but impractical definition of uncertainty, and the dearth of research directly relevant for product development decision makers. Since 2003, both materials and energy prices have been volatile compared to historic price trends. Together with the recent global economic downturn, these events represent a scale of change not seen for decades. Change breeds uncertainty. Despite an increase in public and academic attention, the nature of uncertainty itself remains somewhat elusive. The widely accepted distinction between aleatory (inherently random) and epistemic (related to knowledge) uncertainty is rarely easy to draw or insightful. In addition, a review of the state of the art of Design for Manufacturability, Decision Analysis, and Information Visualization suggests that the assessment of prediction capability represents an element of uncertainty researchers have overlooked.;To address these needs, this research examines the nature of uncertainty, develops and explains the Density Cloud method, investigates the method's impact on engineers making probability assessments, and reviews a case study prototyping these methods. Revisiting the notion of uncertainty within the context of the scientific method suggests that distinguishing between the uncertainty of measurements and the uncertainty of predictions is both feasible and illuminating. In general, the act of prediction must be subjective, perhaps even creative; yet the bulk of published uncertainty-related research addresses measurements or observations. Since decision makers must make use of predictions as well as measurements, the Density Cloud method combines the application of subjective probability theory, a general Monte Carlo method, and a basic data visualization technique in order to express predictions and beliefs both qualitatively and quantitatively. Decision Analytical Scorecarding integrates the Density Cloud method into a the general Decision Analysis framework. To investigate this method, this work summarizes a survey-based experiment comparing numerical probability assessments to visual probability assessments. The experiment confirms prior research investigating the representativeness heuristic, and in particular working engineer's tendency to disregard sample size when numerically assessing probability distributions. Evidently, however, they are more sensitive to sample size when they are presented with visualizations of probability distributions than when asked to numerically assess them. The results also indicate that respondents were no less effective selecting appropriate distributions when presented with density clouds than when presented with histograms I conclude that since visualizations seem to be more intuitive than numbers alone with respect to the representativeness heuristic, and since density clouds are as effective as histograms yet contain more information, that density clouds represent a measurable improvement in visualizing prediction uncertainty over standard visualization techniques.;Further work along these lines might investigate how the Density Cloud method impacts other heuristics and biases, model quality, and decision quality. I have high hopes for the use of this work, particularly in fields that require long-term thinking, such as energy policy, infrastructure development, renewable electricity generation, and many fields related to the sustainability of how we design, produce, deliver, and re-use our products and services. |
